Data Quality for Responsible AI in Energy Systems (DARES)

While powerful AI systems can help achieve ambitious carbon emission reduction goals they are intrinsically non-deterministic and prone to failure. Lack of data quality is one core reason for AI failure, risking blackouts and other energy system failures. For the Nordic Region to become the most sustainable and integrated region in the world, the problem of data quality must be addressed. 
 
The EU AI Act categorises AI systems used in critical infrastructures as high-risk and subject to requirements for data quality. Surprisingly, despite decades of research and standards development, data quality remains an underdefined and inconsistently operationalised concept, in part because data quality is a socio-technical concept, because data are made by people. Policymakers and regulators need clear definitions and an understanding of the socio-technical issues surrounding data quality. Energy industries, from heating to power, recognise data quality as an acute issue and require clear guidelines for creating conditions for high-quality data creation and use. 
 
DARES seeks to address data quality in AI for energy systems through comparative, mixed-method approaches. Leveraging a diverse, interdisciplinary, multi-country consortium, DARES is uniquely positioned to deliver (1) comprehensive policy recommendations for data quality in responsible and sustainable AI development to support Nordic-Baltic green transition goals; (2) well-defined guidelines for data quality goals in data generation processes, AI system development pipelines, and AI system life-cycle for energy industries; (3) software libraries, benchmarks, toolkits, and evaluation approaches for incorporating data quality in synthetic data development and quality aware machine learning approaches; and (4) speculative energy futures workshops for industry and policy makers engaging shared Nordic-Baltic values. 
 
DARES advances the Nordic-Baltic cooperation through intensive knowledge exchange on green transition, data quality, and responsible AI through interdisciplinary research, engagement with regulators and policymakers across all partner countries, and joint participation in speculative envisionings of the Nordic-Baltic energy futures through responsible AI development and implementation.  

At the SVT, Kjetil Rommetveit leads WP1: Mapping policy and regulatory concerns across the Nordic-Baltic region. WP1 will study the regulatory aspects of environmental data quality, focusing on the Nordic-Baltic region and its adaptations to new requirements during the implementation of the EU AI Act. Led by the SVT in cooperation with University of Copenhagen, this WP will rely on document analysis of relevant policy documents, public statements, standards, and reports to map the stated policy intentions and concerns across partner countries. 

The researchers in WP1 will map how AI futures, based in technological potential and promise, are accommodated and made actionable by regulators. Empirical work will focus on how technical and regulatory understandings of AI data are worked out in the Nordic-Baltic settings, including the requirements and obligations of the EU AI Act.